Adaptability in Wireless Sensor Networks Through Cross-Layer Protocol and Architectures

Christophe Merlin

Associate Professor Wendi Heinzelman

CSB 426

Abstract

Complex applications and increased sensor capabilities will help proliferate wireless sensor networks into everyday life. However, as sensor nodes are battery-operated, sensor networks will require protocols to spare every possible bit of energy. This can be accomplished through cross-layer protocol optimizations that specialize the protocols for specific application requirements. However, as researchers continue to contribute to the field of sensor networks, protocols will evolve, and more efficient work will replace older ideas. Therefore, flexibility and eased maintenance of the network will be required to make sensor networks feasible for new deployments and customers. Thus there are competing goals of energy-efficiency, achieved by specialization through cross-layer protocol design, and flexibility, achieved through modularity in a layered protocol design.

My thesis shows that these competing goals can be balanced by the use of cross-layer information exchange that enables the protocols (and hence the network) to adapt to current application and network conditions. Adapting the protocols via cross-layer information exchange allows the network to make best use of the limited energy resources of the sensor nodes while maintaining required application quality of service and retaining a flexible protocol stack. In support of this thesis, we have (1) performed a case-study comparing cross-layer designs with a layered design, (2) designed and evaluated an architecture that enables the sharing of information across traditional stack boundaries, (3) developed various techniques at the MAC layer to take advantage of the availability of this cross-layer information to extend node lifetime, (4) demonstrated the advantage of allowing a sensor network middleware to exploit this cross-layer information to maximize the time when application quality of service is met, and (5) proposed a solution to manage the bidirectional flow of information between middleware and the protocol stack. With the standardization of cross-layer information exchange provided by our new architecture, and the many example protocol adaptations provided in this dissertation, one can envision new designs at all protocol levels, making sensor networks truly adaptive to changes in both application requirements and network conditions.